Use Cases

The Use Case domain defines standardised AI use cases: fixed, named patterns of AI use, each defined on the same scales, so that obligations, assessments and their results compare across jurisdictions, clients and time.

A use case is a reference definition, not a record of a particular organization's system. An organization's own AI systems are Model Risk Management model profiles, and obligations point to the use cases they may reach through useCaseIDs.

Schema: use-case.schema.json
Reference data: reference-data/use-cases.json (UC1 to UC10, validated in CI)
Specification: JSON Schema Draft-07

The standard use cases

#	Use case	Authority	Boundary
UC1	Staff copilot drafts and summarises	A1	B3
UC2	Score or recommendation shown to a decision-maker	A2	B1, B3
UC3	Automated triage or decline	A3	B1, B3
UC4	Automated pricing or straight-through payment	A4	B1
UC5	Agent acting across steps and tools	A5	B1, B3
UC6	Agent with write access to systems of record	A6	B3
UC7	AI inside a provider or the platform behind it	Any	B1, B2
UC8	Customer-facing assistant	A1, A2, A3	B3
UC9	AI-generated code in production systems	A1	B3
UC10	AI tools reachable by a deal team	A1	Channels K1 to K7

Consequence (C1 to C5) is never set on a use case. It depends on the organization's own thresholds, so it is set per matter.

Use Case

Attribute	Type	Description	Required
useCaseID	String	Unique identifier for the use case, e.g. UC3	✓
title	String	Short name of the use case	✓
description	String	What the AI does in this use case, and what it does not	✓
authorityLevels	Array[Enum]	Authority levels the use case spans: how far the AI's output stands in for a human decision or action: a1Productivity, a2Analytical, a3Decision, a4Transactional, a5Agentic, a6Actuation	✓
anyAuthorityLevel	Boolean	True when the use case applies at any authority level (for example AI inside a provider); authorityLevels then lists the levels most often seen
boundaries	Array[Enum]	Boundary positions the use case spans: where the AI sits relative to the organization: b1, b2, b3
confidentialityChannels	Array[Enum]	Confidentiality exposure channels the use case spans, for use cases about access to confidential information: k1, k2, k3, k4, k5, k6, k7
assessmentTypes	Array[Enum]	Kinds of assessment the use case typically leads to: aiRiskAppetite, aiUseAssessment, aiConfidentialityExposureAssessment
examples	Array[String]	Illustrative instances of the use case
exclusions	Array[String]	Patterns that look similar but belong to another use case, with the use case they belong to
relatedUseCaseIDs	Array[String]	Use cases that commonly occur alongside this one
version	String	Version of the use case definition

Scales

The authority scale is defined here. The boundary, consequence, evidence and confidentiality channel scales are defined by the assessment method in use; the schema carries their codes so records and assessments use them consistently.

Authority Level (authorityLevels)

Value	Code	Meaning
a1Productivity	A1	Drafts, summarises or assists a person who remains the author
a2Analytical	A2	Produces a score, analysis or recommendation that a person decides on
a3Decision	A3	Makes or automates a decision, such as triage or decline
a4Transactional	A4	Executes a transaction, such as pricing or a payment, straight through
a5Agentic	A5	Plans and acts across several steps and tools
a6Actuation	A6	Writes to or changes systems of record

Boundary (boundaries)

Value	Code	Meaning
b1	B1	AI in a provider's service used by the organization
b2	B2	AI in the platform behind a provider (a fourth party)
b3	B3	AI the organization deploys within its own boundary for its own use

Assessment Type (assessmentTypes)

Value	Meaning
aiRiskAppetite	Setting the board's AI risk appetite and its triggers
aiUseAssessment	Assessing a particular AI use, including AI in the supply chain
aiConfidentialityExposureAssessment	Assessing exposure of confidential information, such as deal information, to AI tools

{
  "useCaseID": "UC3",
  "title": "Automated triage or decline",
  "description": "AI decides how a case, claim or application is routed, or declines it, without a person deciding each case.",
  "authorityLevels": [
    "a3Decision"
  ],
  "boundaries": [
    "b1",
    "b3"
  ],
  "assessmentTypes": [
    "aiRiskAppetite",
    "aiUseAssessment"
  ],
  "examples": [
    "Automatic decline of a credit application",
    "Claims triage that fast-tracks or refers claims"
  ],
  "version": "1.0"
}